A semi-automated annotation for boar sperm classification using deep convolution neuron network

This study aims to develop a semi-automated annotation based on deep convolution neuron networks in mobility computer-assisted sperm analysis for boar sperm classification (mobile-CASA). The semen from two boar breeders was captured into 250 frames and labeled as good or bad using sperm head morphology. A semi-annotation is split into two processes. At first, an expert reviews the images and annotates a few frames to verify the sperm's annotations. After the expert has finished, relevant frames will be selected and passed on to an AI model recursively until the learned model can accurately identify sperm morphology. The experiment results show that our proposed tend to reduce the workload of the domain expert and improve detection accuracy from 0.65 to 0.95. In addition, this model can embed in mobile devices that are easy to use and access by general farmers.